Flat Field 'auto selection' selects (and averages) raw with mismatching "aspect ratio" (1:1, 4:3, 3:2 etc), giving broken results

[Beep6581]

Originally reported on Google Code with ID 1229

I have a raw image for each aspect ratio and f-stop, in the one folder. The incorrect
raw image is chosen by RT4 for the flat field source (as in, it uses the flat field
file with the wrong aspect ratio). 

An added problem is that RT4 is averaging flat field images (from the same camera/lens)
BUT with different aspect ratios. In my case, that's creating images 3 stops darker
in the right hand third of the image (for 16:9 aspect raws).

Currently using Win Vista and current 32bit release BUT have also tried the latest
source (as of 30/01/2012) on Ubuntu 11.10 and get the same thing.

Camera - DM-GF1 with 20mm 1.7: 

Branch: default
Version: 4.0.6.3
Changeset: 286a06753099
Compiler: GCC 4.6.1
Processor: generic x86
System: Windows
Bit depth: 32 bits
Gtkmm: V2.22.0
Build type: RELEASE
Build flags:  -mtune=generic -O2 -DNDEBUG
Link flags:  -mtune=generic -mwindows -s -Wl,--large-address-aware
OpenMP support: ON
MMAP support: ON

Operating systems winXP and Vista -32bit and Ubuntu-11.10-64bit

PS: rawtherapee is mindblowingly brilliant! so much thanks to all involved.

Thank you.

Reported by kye.wood on 2012-01-31 05:11:33

[Beep6581]

I suppose we will need to add image width and height to the auto matching logic. Could
you please share some raw files, as I don't have access to the multi-aspect ratio raws
from a single camera/lens model.

Reported by michaelezra000 on 2012-01-31 12:21:40

[Beep6581]

This file here raws_with_different_aspect_ratio.zip<http://www.fileserve.com/file/2bewm98/raws_with_different_aspect_ratio.zip>contains
four raw images, from the same Lumix GF1 with the same lens
attached. Each raw was shot in a different aspect ratio. 1:1, 3:2, 4:3 and
16:9.

Good luck with the coding.

Thanks very much. This will be a godsend for me and hopefully others as
well.

Reported by kye.wood on 2012-02-01 08:41:57

[Beep6581]

ok, I will take a look, thanks.

I suppose that exact matching by width and height is sufficient in this case and no
need to account for orientation. 

Fabio & Emil, please let me know if you think otherwise.

Reported by michaelezra000 on 2012-02-01 14:59:51

[Beep6581]

I am basically done with ff matching changes, except not sure how to properly read the
image width and height.

Lumix GF1 has a nice pair of exif tags Sensor Width and Sensor Height, but I don't
see those in other cameras. Image dimensions are correctly reflected in the UI in Crop/W
and H, any hint on how these are populated, I can't seem to trace it.

Should ImageWidth or PixelXDimension be used?
This portion of getting the dimensions would need to be re-coded in the xmp branch.

Reported by michaelezra000 on 2012-02-03 04:47:12

[Beep6581]

Here is the patch that should be able to do additional FF matching by image size.
All is in place except the method of reading the image size itself:) 
So at this point it reads width=0 and height=0 on image to be corrected and all flat
fields, therefore works as before not yet providing any new functionality (very useful,
I know:))

When this is solved the same should probably be done for the dark frame auto selection.

Reported by michaelezra000 on 2012-02-03 18:10:46

---

• Attachment: FF-sensorsize-01.patch

[Beep6581]

I guess someone should compile a Windows build for kye.wood to test.

As for Linux users (or just me?) it fails to compile:

[  0%] Creating the about file
-- hg command found: /usr/bin/hg
[  0%] Built target AboutFile
[  5%] Built target rtexif
[  6%] Building CXX object rtengine/CMakeFiles/rtengine.dir/dfmanager.cc.o
/home/drslony/rawtherapee/rtengine/dfmanager.cc: In member function ‘rtengine::dfInfo*
rtengine::DFManager::addFileInfo(const Glib::ustring&, bool)’:
/home/drslony/rawtherapee/rtengine/dfmanager.cc:274:29: error: cannot declare variable
‘idata’ to be of abstract type ‘rtengine::ImageData’
/home/drslony/rawtherapee/rtengine/imagedata.h:33:40: note:   because the following
virtual functions are pure within ‘rtengine::ImageData’:
/home/drslony/rawtherapee/rtengine/../rtgui/../rtengine/rtengine.h:84:25: note:   
     virtual int rtengine::ImageMetaData::getWidth() const
/home/drslony/rawtherapee/rtengine/../rtgui/../rtengine/rtengine.h:86:25: note:   
     virtual int rtengine::ImageMetaData::getHeight() const
make[2]: *** [rtengine/CMakeFiles/rtengine.dir/dfmanager.cc.o] Error 1
make[1]: *** [rtengine/CMakeFiles/rtengine.dir/all] Error 2
make: *** [all] Error 2

Reported by entertheyoni on 2012-02-10 17:24:48

• Status changed: PatchSubmitted
• Labels added: Component-Quality

[Beep6581]

DrSlony, the patch I attached is a skeleton, it does NOT yet do what is requested as
I cannot get the image dimensions data.

Reported by michaelezra000 on 2012-02-10 17:49:19

[Beep6581]

I try to clean out bug-tracker a bit: What's the status here?

Is it 'WontFix' or is somebody willing to implement this? I'll close it 'WontFix' in
a week if nobody complains.

Reported by rinni@gmx.net on 2013-01-28 22:21:59

[Beep6581]

Reported by rinni@gmx.net on 2013-02-04 20:42:19

• Status changed: WontFix